Patent Application
20200134096
Users can perform searches to retrieve information from a data repository (or multiple data repositories). A data repository can include a database, such as a structured database that is accessed using Structured Query Language (SQL) queries, or a non-structured database. A data repository can be available on a network, such as the Internet, a private network, and so forth.
Some implementations of the present disclosure are described with respect to the following figures.
Throughout the drawings, identical reference numbers designate similar, but not necessarily identical, elements. The figures are not necessarily to scale, and the size of some parts may be exaggerated to more clearly illustrate the example shown. Moreover, the drawings provide examples and/or implementations consistent with the description; however, the description is not limited to the examples and/or implementations provided in the drawings.
In the present disclosure, use of the term “a,” “an”, or “the” is intended to include the plural forms as well, unless the context clearly indicates otherwise. Also, the term “includes,” “including,” “comprises,” “comprising,” “have,” or “having” when used in this disclosure specifies the presence of the stated elements, but do not preclude the presence or addition of other elements.
Search results returned in response to a query submitted to obtain information from a data repository can be based on identification of the search results based on terms included in the query (a term in a query can be referred to as a “search term”). A “term” can refer to a word, a phrase, or any other information that makes up a predicate that indicates what information in the data repository is relevant to the query. For example, a search engine can respond to the query by identifying data records in the data repository that contain terms satisfying (e.g., matching exactly, matching partially, etc.) the search term(s) of the query.
Different users that submit the same search terms in respective queries may expect different results. For example, a member of a finance group of an enterprise seeking information regarding about “tablet computers” (an example of a search term) may seek data records that are different from the data records sought by a member of a technical research and development group of the enterprise using the same search term. The member of the finance group may be interested in documents relating to sales and revenue of tablet computers, while the member of the technical research and development group may be interested in documents relating to recent advancements in technical features of electronic component in tablet computers.
Returning the same collection of documents based on a query containing a given search term regardless of which user submitted the query may produce search results that are not satisfactory for at least some users.
In accordance with some implementations of the present disclosure, documents produced by users (or groups of users) are used to derive models that include terms and respective indications of importance of the terms. A document “produced” by a user or group of users can refer to a document that is created by the user or group, modified by the user or group, or a document having content to which the user or group made a contribution. The models derived based on the documents produced by users or groups of users can be used to identify search results that are more tailored towards interests of users that submit queries for information.
The documents produced by users or groups of users contain terms that are of interest in the day-to-day activities of the users or groups of users. When such terms are considered in performing searches, the likelihood of providing search results more tailored to the interests of the users that submit queries is increased.
As used here, an “engine” can refer to a hardware processing circuit, which can include any or some combination of a microprocessor, a core of a multi-core microprocessor, a microcontroller, a programmable integrated circuit, a programmable gate array, a digital signal processor, or another hardware processing circuit. Alternatively, an “engine” can refer to a combination of a hardware processing circuit and machine-readable instructions (software and/or firmware) executable on the hardware processing circuit.
A data repository 104 can refer to a structured database, such as a database that is accessed using Structured Query Language (SQL) queries. A structured database can store data in relational tables. Alternatively, a data repository can refer to a non-structured database, which stores data in an unstructured manner. In other examples, a data repository can refer to any other collection of information that can be searched by the search engine 102.
Although
A user can use an electronic device to submit a respective search query to the search engine 102. Examples of electronic devices include desktop computers, notebook computers, tablet computers, smartphones, and so forth. Thus, in an example of
The search engine 102 includes targeted search logic 114 according to some implementations of the present disclosure, where the targeted search logic 114 performs a targeted search of data in the data repository 104 using a model that is derived from documents produced by a particular user or a particular group of users.
The targeted search logic 114 can include a portion of the hardware processing circuit of the search engine 102, or alternatively, the targeted search logic 114 can be implemented as machine-readable instructions executable by the search engine 102. In other examples, the targeted search logic 114 can be separate from the search engine 102.
As depicted in
The model 118 is derived based on documents produced by a first group of users, where the first group of users can include the first user 110. By using the model 118 when processing the search query 106 from the first user 110, the targeted search logic 114 is able to return search results that are more targeted towards the interest of the first user 110. Similarly, the model 120 is derived based on documents produced by a second group of users, where the second group of users include the second user 112.
Although
Moreover, although the foregoing examples refer to users submitting search queries to the search engine 102 and models based on documents produced by users or groups of users, it is noted that in other examples, other types of entitles can submit search queries to the search engine 102. Such other types of entities can include machines or programs.
Also, generally, a model used by the targeted search logic 114 can be based on documents produced by an entity or a group of entities, where an entity can refer to any of a user, a machine, or a program.
In the example of
The process 200 receives (at 202) documents produced by a group of entities during operation of the group of entities. As used here, an “operation” of a group of entities refers to an activity (or collection of activities) of a group of entities in performing tasks. The group of entities can collaborate with one another during the operation to collaboratively produce the documents, such as to collaboratively create documents, modify documents, or otherwise make contributions to the content of documents. In examples where the entities are users, a group of users can belong to a department of an enterprise, such as a business concern, an educational organization, a government agency, and so forth. In other examples, other groups of users can be defined. As part of the work of the users, the users can collaborate to produce documents. Examples of documents that can be produced by users include emails, word processing documents, presentations, spreadsheets, summary reports, and so forth.
The process 200 extracts (at 204) terms from the documents produced by the group of entities during the operation of the group of entities. As used here, a “term” can refer to a word, a portion of a word, a phrase that includes multiple words, and so forth. The terms that are extracted can exclude terms that are common terms. For example, words such as “the,” “and,” and so forth are common terms that do not meaningfully aid in producing targeted search results. Such common terms can be referred to as “stop terms,” which are terms that occur with a frequency in documents that is deemed to exceed some frequency threshold.
The process 200 determines (at 206) indications of importance of the extracted terms. The indications of importance can be represented by weights or any other values that provide some indication of the relative importance of an extracted term relative to other extracted terms. In some examples, the indications of importance can be based on a metric produced using a term frequency-inverse document frequency (TF-IDF) technique. In other examples, other term-weighting techniques can be employed.
The weighting factor of the TF-IDF technique is in the form of a TF-IDF value that increases proportionally to the number of times a word appears in a document and is offset (decreased) by the number of documents in a corpus of documents that contain the term. The term frequency (TF) is based on the number of times a term occurs within a document. Thus, a term that occurs more frequently in a document would have a larger TF value. The inverse document frequency (IDF) is based on the number of times the term appears in a corpus of documents. If a term appears in a larger number of documents, then the IDF is large. A larger IDF value offsets the TF value—consequently, a greater frequency of a term in the corpus of documents reduces the overall weighting factor calculated for the term.
The process 200 derives (at 208) a model that includes the indications of importance of the extracted terms. For example, the derived model can include a list of terms and the corresponding indications of importance of the listed terms. In other examples, the model can have a different form.
By using the model derived according to some implementations of the present disclosure, terms that are focused upon by a given group of entities can be identified with greater indications of importance. For example, jargon or domain-specific terms that are used by a group of entities may be terms of particular interest to the group of entities. Members of a finance group may employ different domain-specific terms and jargon as compared to members of a technical research and development group.
The process 200 further receives (at 210) a search from a first entity that is part of the group of entities. In response to the search, the process 200 accesses (at 212) the derived model, to perform a targeted search of a data repository (e.g., 104 in
In performing the targeted search, the targeted search logic 114 determines whether data records of the data repository contain terms that match the terms included in the derived model. If so, the targeted search logic 114 retrieves the respective indications of importance of the terms that match the terms included in the derived model. The indications of importance can be used by the targeted search logic 114 to decide which data records are of higher relevance to the search query. For example, a data record that contains many occurrences of a particular term that is associated with a relatively high indication of importance in the derived model can be identified as being more relevant than another data record that does not contain the particular term or that has a smaller number of occurrences of the particular term.
The machine-readable instructions include model accessing instructions 302 to, in response to a search received from a first entity, access a model derived from documents produced by the first entity or a group of entities comprising the first entity during operation of the first entity or the group of entities, the model comprising indications of importance of terms extracted from the documents.
In some examples, the model is derived from documents produced by the group of entities that collaborate with one another. For example, the group of entities can be part of an enterprise that provides a product or a service. The terms of the model can include jargon words or phrases used by the first entity or the group of entities. As another example, the terms of the model can include terms specific to a domain of the group of entities, such as a finance domain, research and development domain, sales domain, product support domain, etc.
To compute the indications of importance for the model, the machine-readable instructions can extract terms from the documents produced by the first entity or the group of entities during operation of the first entity or the group of entities, count respective numbers of occurrences of the extracted terms (such as numbers of occurrences of the extracted terms in a document or a corpus of documents for driving TF and IDF values as explained above), compute the indications of importance for the extracted terms based on the respective numbers of occurrences of the extracted terms.
The machine-readable instructions further include search result returning instructions 304 to return a search result that is based on the query and on the model.
The system 400 further includes a storage medium 404 that stores machine-readable instructions executable on the hardware processor 402 to perform various tasks. Machine-readable instructions executable on a hardware processor can refer to the instructions executable on a single hardware processor or the instructions executable on multiple hardware processors.
The machine-readable instructions include term extracting instructions 406 to extract terms from documents produced by a group of entities during operation of the group of entities. The machine-readable instructions further include importance indication determining instructions 408 to determine indications of importance of the extracted terms. The machine-readable instructions further include model deriving instructions 410 to derive a model comprising the indications of importance of the extracted terms. For example, the model can include a list of terms and the associated indications of importance (e.g., weights). In further examples, the model can further include information associated with individual entities (e.g., user identifiers or identifiers of other entities, location information of entities, etc.) and information associated with the group of entities (e.g., group identifiers, location information of a group, etc.).
The machine-readable instructions also include model accessing instructions 412 to, in response to a search received from a first entity that is part of the group of entities, access the model. The machine-readable instructions further include search result returning instructions 414 to return a search result that is based on the query and on the model.
The storage medium 300 (
In the foregoing description, numerous details are set forth to provide an understanding of the subject disclosed herein. However, implementations may be practiced without some of these details. Other implementations may include modifications and variations from the details discussed above. It is intended that the appended claims cover such modifications and variations.
